ironic.conductor.manager module

Conduct all activity related to bare-metal deployments.

A single instance of ironic.conductor.manager.ConductorManager is created within the ironic-conductor process, and is responsible for performing all actions on bare metal resources (Chassis, Nodes, and Ports). Commands are received via RPCs. The conductor service also performs periodic tasks, eg. to monitor the status of active deployments.

Drivers are loaded via entrypoints by the ironic.common.driver_factory class. Each driver is instantiated only once, when the ConductorManager service starts. In this way, a single ConductorManager may use multiple drivers, and manage heterogeneous hardware.

When multiple ConductorManager are run on different hosts, they are all active and cooperatively manage all nodes in the deployment. Nodes are locked by each conductor when performing actions which change the state of that node; these locks are represented by the ironic.conductor.task_manager.TaskManager class.

A tooz.hashring.HashRing is used to distribute nodes across the set of active conductors which support each node’s driver. Rebalancing this ring can trigger various actions by each conductor, such as building or tearing down the TFTP environment for a node, notifying Neutron of a change, etc.

class ironic.conductor.manager.ConductorManager(host, topic)[source]

Bases: ironic.conductor.base_manager.BaseConductorManager

Ironic Conductor manager main class.

RPC_API_VERSION = '1.50'
add_node_traits(**kwargs)
change_node_power_state(**kwargs)
continue_node_clean(context, node_id)[source]

RPC method to continue cleaning a node.

This is useful for cleaning tasks that are async. When they complete, they call back via RPC, a new worker and lock are set up, and cleaning continues. This can also be used to resume cleaning on take_over.

Parameters
  • context – an admin context.

  • node_id – the id or uuid of a node.

Raises

InvalidStateRequested if the node is not in CLEANWAIT state

Raises

NoFreeConductorWorker when there is no free worker to start async task

Raises

NodeLocked if node is locked by another conductor.

Raises

NodeNotFound if the node no longer appears in the database

continue_node_deploy(context, node_id)[source]

RPC method to continue deploying a node.

This is useful for deploying tasks that are async. When they complete, they call back via RPC, a new worker and lock are set up, and deploying continues. This can also be used to resume deploying on take_over.

Parameters
  • context – an admin context.

  • node_id – the ID or UUID of a node.

Raises

InvalidStateRequested if the node is not in DEPLOYWAIT state

Raises

NoFreeConductorWorker when there is no free worker to start async task

Raises

NodeLocked if node is locked by another conductor.

Raises

NodeNotFound if the node no longer appears in the database

create_allocation(**kwargs)
create_node(**kwargs)
create_port(**kwargs)
destroy_allocation(**kwargs)
destroy_node(**kwargs)
destroy_port(**kwargs)
destroy_portgroup(**kwargs)
destroy_volume_connector(**kwargs)
destroy_volume_target(**kwargs)
do_node_clean(**kwargs)
do_node_deploy(**kwargs)
do_node_rescue(**kwargs)
do_node_tear_down(**kwargs)
do_node_unrescue(**kwargs)
do_provisioning_action(**kwargs)
driver_vendor_passthru(**kwargs)
get_boot_device(**kwargs)
get_console_information(**kwargs)
get_driver_properties(**kwargs)
get_driver_vendor_passthru_methods(**kwargs)
get_indicator_state(**kwargs)
get_node_vendor_passthru_methods(**kwargs)
get_node_with_token(**kwargs)
get_raid_logical_disk_properties(**kwargs)
get_supported_boot_devices(**kwargs)
get_supported_indicators(**kwargs)
heartbeat(**kwargs)
inject_nmi(**kwargs)
inspect_hardware(**kwargs)
object_action(context, objinst, objmethod, args, kwargs)[source]

Perform an action on a VersionedObject instance.

Parameters
  • context – The context within which to perform the action

  • objinst – The object instance on which to perform the action

  • objmethod – The name of the action method to call

  • args – The positional arguments to the action method

  • kwargs – The keyword arguments to the action method

Returns

A tuple with the updates made to the object and the result of the action method

object_backport_versions(context, objinst, object_versions)[source]

Perform a backport of an object instance.

The default behavior of the base VersionedObjectSerializer, upon receiving an object with a version newer than what is in the local registry, is to call this method to request a backport of the object.

Parameters
  • context – The context within which to perform the backport

  • objinst – An instance of a VersionedObject to be backported

  • object_versions – A dict of {objname: version} mappings

Returns

The downgraded instance of objinst

object_class_action_versions(context, objname, objmethod, object_versions, args, kwargs)[source]

Perform an action on a VersionedObject class.

Parameters
  • context – The context within which to perform the action

  • objname – The registry name of the object

  • objmethod – The name of the action method to call

  • object_versions – A dict of {objname: version} mappings

  • args – The positional arguments to the action method

  • kwargs – The keyword arguments to the action method

Returns

The result of the action method, which may (or may not) be an instance of the implementing VersionedObject class.

remove_node_traits(**kwargs)
set_boot_device(**kwargs)
set_console_mode(**kwargs)
set_indicator_state(**kwargs)
set_target_raid_config(**kwargs)
target = <Target version=1.50>
update_node(**kwargs)
update_port(**kwargs)
update_portgroup(**kwargs)
update_volume_connector(**kwargs)
update_volume_target(**kwargs)
validate_driver_interfaces(**kwargs)
vendor_passthru(**kwargs)
vif_attach(**kwargs)
vif_detach(**kwargs)
vif_list(**kwargs)
ironic.conductor.manager.do_sync_power_state(task, count)[source]

Sync the power state for this node, incrementing the counter on failure.

When the limit of power_state_sync_max_retries is reached, the node is put into maintenance mode and the error recorded.

Parameters
  • task – a TaskManager instance

  • count – number of times this node has previously failed a sync

Raises

NodeLocked if unable to upgrade task lock to an exclusive one

Returns

Count of failed attempts. On success, the counter is set to 0. On failure, the count is incremented by one

ironic.conductor.manager.get_vendor_passthru_metadata(route_dict)[source]
ironic.conductor.manager.handle_sync_power_state_max_retries_exceeded(task, actual_power_state, exception=None)[source]

Handles power state sync exceeding the max retries.

When synchronizing the power state between a node and the DB has exceeded the maximum number of retries, change the DB power state to be the actual node power state and place the node in maintenance.

Parameters
  • task – a TaskManager instance with an exclusive lock

  • actual_power_state – the actual power state of the node; a power state from ironic.common.states

  • exception – the exception object that caused the sync power state to fail, if present.